Occupied focus inversion about hanabi.github.io HOT 52 CLOSED

actually i guess we can just word it such that "OFI only applies if it is impossible to be a more-ordinary OPC or an occupied finesse"

from hanabi.github.io.

Interesting idea. I like it. 2 thoughts.

1. Wording clarify for situations with 3 cards from 'other' to the 'next card' (it should be implied that this means left most for chop focuses clues, and 2nd left most for left focused clues.

2. My main concern is to caution against out cases where Romain's playable is not globally known.
   This is the same issue that plagues priority and a common intermediate mistake.

from hanabi.github.io.

I'm confused. Who's Alice in this example? What other card? There's only one card clued.

from hanabi.github.io.

Romain is Alice with a yellow 1 to play.
Instead of playing y1 Romain clues yellow to asaelr.

The fact that both are yellow is not related here. What is important is asaelr sees no reason why Romain should give this clue instead of play knowing Kakashi could give the clue instead.

from hanabi.github.io.

seems fine

from hanabi.github.io.

Example 2)

I have a known r4 and a known b3 (b5 finesse).
Asaelr know it was a finesse since i have a b3.
I can clue 3 to Valetta.

edit: Mmm, when it's used with number, it conflict with a single bluff, so allow it with number is way less powerfull.

from hanabi.github.io.

Would this also work to change focus on, say, a clue that hits slots 2 and 3? Or only clues that would normally target chop?

from hanabi.github.io.

Some further thoughts from observing the second example.

Is it illegal if Romain's b3 IS a blind play?

• as yet undefined as you could arrange the priority of conventions as you want, but most likely no, it would still be allowed. Asaelr can see the clue to Valetta works as a OPC focus inversion and considers that the intended outcome (as opposed to a stacked finesse for his b3 and g2). This is a good point for debate though.

Is it illegal if Romains b3 is blind and the focus inversion is on asaelr?

• almost certainly yes. Asaelr will be unable to untangle whether the b3 is gotten and the clue is a focus inversion vs the b3 was ambiguous and the clue is a normal clue to him. As he is not in a position to reasonably entertain both options, one must be prescribed as illegal to prevent confusion and/or bad tempo.

from hanabi.github.io.

Actually, the more i look at it the more i seem inclined to agree with Romain's comment that the conflict with Bluff is too much to ignore.

Option a) Allow OPC Focus inversion to take priority over bluffs and finesses. So IF the player is ever occupied and gives a bluff or finesse, IF the second touched card is playable you must assume as such.

Option b) OPC focus inversion requires that both the focus and inversion card are the same suit. Thus from good touch principle the bluff interpretation is illogical. Or that the focus is Known trash (eg clue two 3s but only 1 3 could be good as rest are all played/touched already)

I think i prefer B. It is certainly the smaller change, i think it's the better one (but not sure).

from hanabi.github.io.

So does Option a) allow finesse on the focus inversion? Like on the second example change all the 3 into 4 in Valetta's hand, and change g4 into y2.

Does option a) allow bluff? Change all the 3 into 4 in Valetta's hand, add a purple suit with 3 played, and remplace g4 by p4 (or hard bluff, remplace g4 by k1).

I feel like it's easier than assume OPC focus inversion only have priority if the focus inversion is playable.
You will always assume it's a focus inversion with that way of thinking.

from hanabi.github.io.

Would this also work to change focus on, say, a clue that hits slots 2 and 3? Or only clues that would normally target chop?

Yes, I was searching for an example of that.
But most of the (limit or bad) example I found are in early game, so you can generally just make a player not occupied do the clue with color once the card arrive on chop. So (I think) for the clue to be worth it, you really want something extra, like an useless card on the player after you who could give the same clue, or an usefull and not critical card on the player who receive the clue.

from hanabi.github.io.

I have a known r4 and a known b3 (b5 finesse).
Asaelr know it was a finesse since i have a b3.
I can clue 3 to Valetta.

This conflicts with the move called the Occupied Finesse.
Currently, the way that we play is if Romain clues 3's to Valetta, Valetta should blind-play her Finesse Position as a Occupied Known Self-Finesse (meaning red 2 into red 3 or green 2 into green 3). Valetta would NOT assume Focus Inversion and, after blind-playing a 2, would go on to play her chop 3 from Chop Focus.

Is the intent to delete this kind of Occupied Finesse and replace it with Focus Inversion? If yes, then I don't currently feel that is a very good configuration. Occupied Finesse are more efficient than Focus Inversion. And they are quite common too!

Thus, I think that this convention should only be allowed with color. (Is this more or less what Mercy is saying in his option B above?)

Is it illegal if Romain's b3 IS a blind play?

I think it should be illegal if Romain has any kind of blind-play to do. Otherwise it looks like an Ambiguous Finesse, etc. As an aside, one of the most infuriating things is when players tell me in a post-game review that they did not blind-play into a Finesse because they were trying to perform a Priority Finesse instead. And trying to do Focus Inversion instead of blind-playing into a Finesse is more of the same bullshit.

Would this also work to change focus on, say, a clue that hits slots 2 and 3? Or only clues that would normally target chop?

Yes. How Focus Inversion works is that, in a clue where slots 2 and 3 are hit, if slot 2 was normally the focus, then it would invert to slot 3 (and vice versa).

from hanabi.github.io.

It's partially what I was saying for b), but not exactly.

In multicolor variants color clues on that touch different colored cards (eg a rainbow and yellow card) are still not allowed.
Only in situations where finesse or bluff are not possible is is allowed. This includes

• color clues on cards of the same color so focus inversion is self evident from good touch
• number (or rarely color) clues where the clue focus is known trash after focus inversion occurs.

To put it another way:
a) if it could be a focus inversion OPC it is OPC. Otherwise it is bluff.
b) if it could be a finesse or bluff, its a finesse or bluff. Otherwise it is OPC.

from hanabi.github.io.

ok we so agree that normal OPC interpretations should take priority over OFI
and we agree that color clues on cards of the same color is a clear OFI

number (or rarely color) clues where the clue focus is known trash after focus inversion occurs.

can you type out an alice and bob story where Alice uses a number clue to initiate OFI and it does not conflict with OPC

from hanabi.github.io.

Example 1: Mid game of a rainbow 6 suit game, 4 players.

• All 1s have been played, all 2s have been played except the b2.
• Alice has a playable r4. Bob has nothing to do.
• Cathy's hand is [X m2 b2 X] all unclued. Her hand has been static pretty much all game.
• Alice invokes OFI and clues Cathy 2s. She sees a b3 in Donalds hand and doesnt want to lose tempo on the blue suit, waiting/hoping for Cathy to finally discard. She also feels the r4 playing right now is pretty irrelevant and OFI is much clearer than trying to invoke a stale 2s context read.

This is the 'trash example' which relies on the fact that the clue can never be a finesse or bluff, simply because the conventional clue focused card will never be playable. Note that my example is more contrived than necessary. So long as the focused 2 is trash everyone except Cathy knows the clue is OFI. And Cathy knows the clue is unusual in some way due to Alice giving it over Bob, with OFI being the best logical conclusion.

BAD EXAMPLE - left for posterity.
Example 2: No variant 5 suit 4 player. Board played is b1 r3 y0 g1 p0

• Alice has r4 to play again. It is globally known.
• Bob has nothing to do. His hand has a y5 saved with yellow (5CE) and 3 unknown cards that are negative yellow cards.
• Cathy's hand is [y3 b3 X X].
• Donalds hand is also void of interest.
• Alice clues 3s to Cathy. Due to the unique nature of both Bob's hand's negative information he is able to rule out any chance that this clue is intended to focus on the y3 and so It must be OFI. INCORRECT he assumes it is a 3 bluff*** (there maybe a case where Bob knows it's never on him, this is not valid though). Cathy sees that Alice's choice to clue her was strange when Bob could. OPC is irrelevant again as the clue is never a save. There are no finesse or reverse finesse options. Only OFI remains, and Cathys plays the b3.
  Donald Sees Cathy play the wrong 3 and is surprised. He had marked his hand for y1 and y2 after Alice's original clue. He checks to see if there is context that the b3 played from good touch principle and cannot see anything simple. Thus he assumes the clue is OFI and discards instead of playing slot 1.

Example 3 - BAD: No Variant 4 player Game. Board played is b1 r3 y0 g1 p0

• Alice has nothing to do.
• Bob has globally known r4 to play
• Cathy hand is [y3 b3 b2 X].
• Donalds hand has nothing to do.
• Alice discards seeing no clues to give ending the early game. Bob then clues Cathy 3s. Cathy is surprised Bob is giving a play clue instead of playing, after Alice discarded but knows Alice couldve been invoking the finesse position exception and misplays her y3. OFI Fails again.

It seems to me there it is quite difficult , perhaps impossible to give an example where both 3s are good, without giving OFI some priority over other existing conventions. We could do this, im leaning towards no.

from hanabi.github.io.

are you okay with leaving the number case undocumented since it is so rare

from hanabi.github.io.

One game, three screen.
Scharkbite have a nice y2+y4 we want to get, unfortunately, k2 is blocking the clue.

Example 3)

I suppose it didn't work, since I can't do the clue myself because i have y1, it seem for me logical to consider it a yellow clue now would mean to play y4.

Example 4) one turn later

Now it's my turn, and again, I don't think the clue would trigger the OFI, because asaelr perhaps consider his chop less usefull than mine, so he want me to clue that y2 in slot 5.

Example 5) five turn later

Now again, there is a finesse on g2+g3+g4 working. But I have no play. Asaelr have a play, and Scharkbite know Asaelr have a play.
So it should trigger here.
(but again, you was not fan about delaying blind plays, even if it's one turn later)

from hanabi.github.io.

I assume at all points the OFI clue you are referring to in your example above is yellow. If so i agree with your evaluation with a few comments

• The upside to the clue never seems huge here. At best it saves a y4 which is unique, bu also the least valuable card in scharkbite's hand.
• There is no risk here of the 'blind play delay' violation because:
  a) Primarily the 2 is touched. Good touch principle protects us knowing which g2 is called. there is no ambiguous prompt conventions.
  b) practically in 3 player you are almost always safe from ambiguous clues. in this case scharkbite 100% knows his g4, and noone else exists to think they have g2 and/or g3.

So i agree the clue works here exactly as you describe (and only at the end

from hanabi.github.io.

Example 3)
I suppose it didn't work, since I can't do the clue myself because i have y1, it seem for me logical to consider it a yellow clue now would mean to play y4.

• normally it looks like a save on y2, which you have to do before playing the y1
• however, shark will save the b2, so a save on y2 is unnecessary
• so it probably still works as OFI

Example 4) one turn later
Now it's my turn, and again, I don't think the clue would trigger the OFI, because asaelr perhaps consider his chop less usefull than mine, so he want me to clue that y2 in slot 5.

• I think it would trigger OFI for the same reasons as in example 3
• it is very bad teamwork for asaelr to make you get the y2, i think that is absurd and players should not entertain that possibility

from hanabi.github.io.

Occupied Focus Inversion

• Normally, when a player is Occupied and they give a clue to someone else, it signifies one of the following three situations:
  1. They are giving a Chop Focus clue to the next person as an urgent Play Clue or Save Clue before it gets discarded.
  2. They are giving a clue that would normally look like a Save Clue with the intent to transform it into a Play Clue (as an Occupied Play Clue).
  3. They are giving a clue that would normally look like a 1-for-1 Play Clue with the intent to transform it into a 2-for-1 Self-Finesse (as an Occupied Self-Finesse).
• If none of the above situations apply and an Occupied player gives a Play Clue that touches two or more brand new cards, then they must be trying to communicate something extra.
• In this situation, they are trying to communicate that the focus of the clue should be inverted. The Play Clue is either on the left-most card (if it would normally be Chop Focus) or on the right-most card (if it would normally be the left-most card).
• Occupied Focus Inversion takes precedence over an Out-of-Order Finesse. (This follows from Occam's Razor; if it could just be a normal Focus Inversion play clue, then it does not necessarily promise any blind-plays.)
• For example, in a 4-player game:
  • It is the first turn and nothing is played on the stacks.
  • Alice clues red to Cathy, touching a red 1 as a Play Clue.
  • Bob clues blue to Alice as a Play Clue on a blue 1.
  • Instead of playing her red 1, Cathy clues red to Donald, touching a red 3 on slot 2 and a red 2 on slot 3. This must be a Play Clue, since these cards are not on Donald's chop.
  • Donald knows that since Cathy is not playing her red 1 and giving a clue, she must have a good reason.
  • Donald looks around the table and does not see the red 2 in anybody else's Finesse Position. Thus, this cannot be a Selfish Finesse.
  • Donald also knows that this cannot be a Self Finesse or Self Bluff. (Self Bluffs with color are explicitly illegal.)
  • This must be an Occupied Focus Inversion, so Donald marks his right-most card as red 2 for later. The other red card in his hand can be red 3, red 4, or red 5. Donald discards.

from hanabi.github.io.

Seem fine.
You didn't inclued a 3 card clue with like r3-r2-r5-x in Donald's hand because we got no example of it?

from hanabi.github.io.

But we are in early game on your example.
So when Bob discard he said to the team there is no play clue to give.
So when Cathy clue Donald, it should be focus inversion because of focus inversion convention and not because of occupied focus inversion convention, right?

from hanabi.github.io.

In your example, Bob can neither clue 3 to Donald as a self finesse nor 2 to Donald as a play clue? If we’re saying both these things hold, it might be worth specifying Donald’s exact hand that caused this to be possible.

from hanabi.github.io.

So when Cathy clue Donald, it should be focus inversion because of focus inversion convention and not because of occupied focus inversion convention, right?

yes i will rewrite the example

from hanabi.github.io.

@jakestiles I will wait to have an example of it with 3 cards, for seeing what we can do with it.
Perhaps it's too complicated.

from hanabi.github.io.

If you change the action of Bob by something else unrelated (clueing a color on a 1 or clueing 5 as a save clue) it should be fine Zamiell.

from hanabi.github.io.

oh ok

from hanabi.github.io.

what about this:

Bob clues 5 to Alice as a 5 Save.

from hanabi.github.io.

Mmm...
Do you prefer perfect example or clear example?

Because I still see two problems :

• Bob didn't clue r2. It could be blocked. Perhaps he planne an OFI, but it stay strange
• Bob didn't clue 3 to Donald as a finesse. I suppose Donald is in bluff seat so is either forced to save Alice's 5 and assume a finesse | or trust Bob that Alice will have something to do on her turn.

from hanabi.github.io.

you can write a new example?

from hanabi.github.io.

At the middle, remplace by :
"Bob clues blue on b1 to Alice as a play.
Instead of playing her red 1, Cathy clues red to Donald, touching a red 3 on slot 2 and a red 2 on slot 3. This must be a Play Clue, since these cards are not on Donald's chop."

from hanabi.github.io.

ok i updated the example, you can re-check it

from hanabi.github.io.

If it’s a red 4 or 5 instead of r3, does it become OOO or does it work exactly the same way?

from hanabi.github.io.

it would be OOO edit - it would work exactly the same way

from hanabi.github.io.

from hanabi.github.io.

OFI only applies as a last resort
in general, that is typically how we implement new conventions, e.g. that they only apply if it can't be one of our existing conventions

from hanabi.github.io.

So that means not only does r2 get marked, r3 should be marked and playable as well.

no, the r3 could be r5, its not a double play clue, its only focus inversion

from hanabi.github.io.

i've edited the above example to clarify

from hanabi.github.io.

So if it was r4 or r5, it would be OOO? If Cathy sees no r3 on other people’s finesse, Cathy can rule out r4 or r5 and mark r3?

from hanabi.github.io.

i dont understand your question. how could it be a finesse on red 3 if nobody has the red 2

from hanabi.github.io.

why dont you type up an alice and bob story

from hanabi.github.io.

Here’s my updated story:

This must be an Occupied Focus Inversion, so Donald marks his right-most card as red 2 for later. Furthermore, if the other red card in his hand is red 4 or red 5 then this is a OOO finesse calling for r3 (and maybe r4). If Donald sees no r3 on finesse position, Donald can also immediately mark their other red card as r3. Donald discards.

from hanabi.github.io.

no, thats not how it works

from hanabi.github.io.

i added the following to the above example, which was implied before but is now explicit:

• Donald looks around the table and does not see the red 2 in anybody else's Finesse Position. Thus, this cannot be a Selfish Finesse.

from hanabi.github.io.

additionally, see #97

from hanabi.github.io.

So you’re saying Donald has to entertain leftmost r5 and rightmost r2? Even though Donald knows Alice and Bob would interpret this as a OOO finesse and would be promised r3+r4 which Donald knows they don’t have.

from hanabi.github.io.

oh i see what you are saying

from hanabi.github.io.

it would be OOO
OFI only applies as a last resort

i think what i said above is wrong,
OFI must take priority over OOO in the case where the occupied player is the one initiating the OOO. (otherwise i guess OFI would not exist at all and it would just always be OOO, right?)

from hanabi.github.io.

Example 6)
Brown Up & Down, so brown on chop is a play clue.

Valetta clue brown to Libster

from hanabi.github.io.

i added the following to the above example:

• Occupied Focus Inversion takes precedence over an Out-of-Order Finesse. (This follows from Occam's Razor; if it could just be a normal Focus Inversion play clue, then it does not necessarily promise any blind-plays.)

from hanabi.github.io.

/accept

from hanabi.github.io.

• Some time has passed since this issue was opened and the group appears to have reached a consensus.
• ✔️ This change will be integrated into the official reference document.
• This issue will now be closed. If you feel this was an error, feel free to continue the discussion and a moderator will re-open the issue.

(For more information on how consensus is determined, please read the Convention Changes document.)

from hanabi.github.io.